Retracted Article: Track recognition algorithm based on neural network for rail transit

Face recognition technology is an important branch based on biometrics technology, and has broad application prospects in the fields of law, business and security. The purpose of this paper is to propose a neural network-based orbital traffic face recognition algorithm to solve the problem of high traffic volume, high-dimensional and small samples in face recognition, and classic face recognition convolutional neural network. In the (CNN) structure, due to the deep design of the network hierarchy, the calculation is large and the training takes a long time. This paper proposes that the algorithm can effectively reduce the training time and increase the training recognition accuracy. In this paper, the CAS-PEAL face database is used as an experiment, and the face image is preprocessed first. Grayscale, median filtering, edge enhancement and normalization processing are performed on the detected face; secondly, the feature extraction is performed on the face by PCA, and the calculated feature vector is used for the next network training, after elastic momentum The weight adjustment method is completed by training the BP neural network and using the extracted data for the prediction of the face recognition neural network classifier. Finally, the experimental results on the CAS-PEAL face database show that the recognition rate reaches 93.5%. The network structure can reduce the training time and effectively improve the accuracy of face recognition.

Перевод пока недоступен